noshade.eth

EVMaverick #314 | Son of the Sun

[noshade.eth @0xNoShade]

@ForwardEditor @athyuttamre It’s really hard to tell what context it has, I often find I need to remind it of threads we have going, or things earlier in our conversation… also can’t tell what context it’s retained when I start voice up mode again

[noshade.eth @0xNoShade]

@SarahRosenstei3 @Hitchslap1 Yeah but… (up to a point) people are actually drawn to success- not intelligence, and (up to a point) intelligence doesn’t correlate with success

[noshade.eth @0xNoShade]

@DCinvestor Did you play WoW? MMO Analogy: They’re teaching you how to play the end game (e.g. run dungeons), but you’re a noob that realises min/maxing rn is better… and when you’re where you want to be (max level | strong & fit) you’ll probably start playing like that

[noshade.eth @0xNoShade]

@TrustlessState @0xcoinn Fake diamonds... so cubic zirconia?

[noshade.eth @0xNoShade]

@dnotreply @ianzelbo Only 60hz :(

[noshade.eth @0xNoShade]

@shadcn There’s this strange juxta between “I can build everything now!” causing burnout, and the apathy of “If I just wait, I can one shot this on the next model”

Stuck feeling like I need to be building all the time but everything I’m building is going to be pointless

[noshade.eth @0xNoShade]

@joaomendoncaaaa @banteg @Crypto_McKenna Because that’s what Anthropic said…

Are you implying that it’s not better than available models at vulns? or that, even if it is, it doesn’t matter?

[noshade.eth @0xNoShade]

@thekitze kinda cool that my life’s gonna be a sci-fi movie though… and I don’t know how it ends

[noshade.eth @0xNoShade]

@banteg @Crypto_McKenna do you think it’s more ethical ethical to make the model publicly available today at equivalent/premium prices than to do this “Project Glasswing” thing?

If it’s true, I’m pretty sure releasing it would result in billions of damages- blackmail galore, b2b software destroyed, etc

[noshade.eth @0xNoShade]

@max_paperclips Feels too early to be compressing the language to save a few tokens, gonna let the generative wikis settle a bit before adding something like that

[noshade.eth @0xNoShade]

@jlongster Yeah I’m building something to connect different curated knowledge-bases/wikis to your agent depending on what context you’re working with

Each knowledge-base is compressed (constrained to token budget) into a summary that’s kept in the agent’s context

https://t.co/aVLcM81jVQ

[Andrej Karpathy @karpathy]

Wow, this tweet went very viral!

I wanted share a possibly slightly improved version of the tweet in an "idea file". The idea of the idea file is that in this era of LLM agents, there is less of a point/need of sharing the specific code/app, you just share the idea, then the other person's agent customizes & builds it for your specific needs.

So here's the idea in a gist format: https://t.co/NlAfEJjtJV

You can give this to your agent and it can build you your own LLM wiki and guide you on how to use it etc. It's intentionally kept a little bit abstract/vague because there are so many directions to take this in. And ofc, people can adjust the idea or contribute their own in the Discussion which is cool.

[noshade.eth @0xNoShade]

@8teAPi It’s in the set now though- they’re gonna test the next frontier model with OpenClaw before they make it public

[noshade.eth @0xNoShade]

@burkeholland i guess we can create 1,000's of audited deterministic scripts and make them work within that

but feels like there's a "bitter lesson" still, unbounded agents will always outperform, so that's where the forefront will be

[noshade.eth @0xNoShade]

@burkeholland you audit the agent's curated markdown so it's context is "deterministic"

[noshade.eth @0xNoShade]

There’s a compounding effect where a model sycophants your idea, you go on X, engage with content from people in the same boat, and then your feed starts sycophanting you too

[noshade.eth @0xNoShade]

@jcarlossoto @openclaw @NousResearch @steipete do I keep seeing this shit cause it’s true, or is it cause I engage with it? I can’t even tell now- damn algorithm

[noshade.eth @0xNoShade]

@toly What’s left as a meaningful distinction of “A.G.I.” at that point?

It needs to be Quantum and breaking encryption— otherwise it’s just a dumb business running, science discovering, art creating, algorithm building, stock investing agent?

[noshade.eth @0xNoShade]

@toly Have you worked with agent that’s controlling your computer, looking at the screen, opening apps, trying things, testing them, iterating, etc?

These will absolutely get way better over the next 24 months, and, without quantum computing, they will be capable of virtually anything

[noshade.eth @0xNoShade]

@sebstaq @kr0der You can tell the agent “you’re using too many words and I’m getting annoyed about it”

[noshade.eth @0xNoShade]

@benhylak Were you at Apple when GPT-5.4 and Claude-4.6 were available? They were a step change for agents

I'm sure there's lots of problems with "KAIROS"-like systems today, but it seems obvious that frontier models will ace it soon enough